xthexder

After a little more thinking about the problem you’re trying to solve, an op-amp voltage buffer with an off the shelf DC-DC converter for the high voltage rail might be the simplest overall circuit. Unfortunately the ATtiny you’re using doesn’t have a digital to analog (DAC) output, so it’s hard to set the voltage reference with that setup. You could probably rig up something with a PWM channel and a capacitor on the output, and adjust the gain on the op-amp feedback so you can reach the max output voltage you want (6V ?)

Unless of course the goal is actually to learn how to build your own variable output buck-boost converter, in which case, all the power to you!

Ah okay, I read a stat for max total current on a pin group, but I guess the single pin current is lower.

If you have a capacitor on your supply pin, it should reduce fluctuations due to changing outputs, but I found the stat you’re talking about: At 5.0V supply a 15mA load only guarantees 4.0V.

So for it to be reliable you’ll need to use another driver in-between. A single smaller transistor with a fast switching time and low voltage drop when saturated/on would be able to multiply that 5-10mA up to the 100mA you might want, and stay above the 4.5V threshold. You can probably find something within 0.4V.

In terms of switching time, I think the main thing is just picking a frequency outside the audible range, so something like 30kHz or above would work. You can probably figure out your own tradeoffs going to a higher frequency. I think the main benefit is you can use smaller capacitors and get less ripple on the output, but it makes the switching control harder to get right.

The ATtiny datasheet has a lot of useful information about power dissipation for its IO pins. The input capacitance of your MOSFET gate seems to negligible effect the power usage compared to the 30 Ohm gate resistance. At 5.5V that’s 183mA and above the 100mA supply limit, so you’d only actually need a minimum of a 25 Ohm resistor (+30Ohm gate) in series (plus maybe a heatsink depending on your duty cycle).

If you can handle supplying 100mA, then you’d get roughly a 60ns rise time based on the 1nF capacitance, which should be plenty fast for switching anything under 100kHz. This scales linearly, so selecting a resistance to target 10mA would give you a 600ns rise time (thus lowering the switching frequency you can hit before overheating the MOSFET).

It looks like it might be possible. Someone’s done it before on a VIC-20
https://github.com/Kweepa/vicdoom